In the Crane Index™ scan of 1 October 2026, 2,295 websites across 21 industry benchmarks were run through 45 checks of how well AI systems can read them. 54% had no Organisation markup, the code that tells a machine who a business is. 93% named no author in a form a machine can read, and 97% described nothing they sell in structured data.
3,368 websites were approached and 1,073 could not be read at all, so the figures below cover the 2,295 that could. One page was read per site. 105 organisations sit in two benchmarks and are counted in each, so the 2,295 scans cover 2,190 distinct websites.
A site that could not be read is unmeasured, not failed; those sites are covered in AI crawler access 2026.
Key statistics, 2026
Each figure links to itself, so it can be cited on its own.
- 54%1,237 of 2,295 websites (54%) had no Organisation, LocalBusiness or Person markup. Link
- 42%965 (42%) carried no JSON-LD structured data at all. Link
- 64%1,470 (64%) had no sameAs links, the links that let a machine confirm a business against its other profiles. Link
- 93%2,128 (93%) had no Person or author markup; 49 (2%) named an author with credentials a machine could read. Link
- 97%2,234 (97%) had no Product, Service or Offer markup for what they sell. Link
- 83%1,906 (83%) had no breadcrumb markup. Link
- 74%1,688 (74%) offered no machine-readable action, such as a bookable service or a priced product an agent could act on. Link
- 10%229 (10%) were built around the questions a buyer asks. Link
- 25%584 (25%) had an llms.txt file at the root. llms.txt adoption by industry. Link
- 88 of 2,295 (under 1%) backed their claims with links to sources a machine can follow. Link
- 138138 (6%) served almost no content until JavaScript ran, so a reader that does not run it sees close to nothing. Link
- 8686 (4%) put an interactive CAPTCHA in front of the page, which an automated agent cannot pass. Link
- 2525 (1%) blocked an AI answer crawler in robots.txt. AI crawler access in full. Link
Every check, counted
The score has three parts, and each part is a set of checks. Pass means the scanner found what the check looks for. Partial means it found some of it, or that the check is forward-looking and records an absence as partial rather than a failure. Fail means it found none of it. How each check is weighted is set out in the methodology.
AI Search Readiness
Can the engines that read live sites resolve the organisation and extract what it does and sells.
| Check | Pass | Partial | Fail |
|---|---|---|---|
| Entity & Trust | |||
| Organisation / entity schema | 1,05846% | 00% | 1,23754% |
| Entity disambiguation (sameAs) | 70631% | 1195% | 1,47064% |
| Logo & contact in schema | 53323% | 49422% | 1,26855% |
| Authorship & credibility | 492% | 1185% | 2,12893% |
| Breadcrumbs | 38917% | 00% | 1,90683% |
| Entity detail (description & expertise) | 1697% | 26812% | 1,85881% |
| Social proof schema | 311% | 2,26499% | 00% |
| Structured Data Depth | |||
| Valid structured data (JSON-LD) | 1,30457% | 261% | 96542% |
| Linked entity graph | 55424% | 31714% | 1,42462% |
| Rich content types | 1858% | 00% | 2,11092% |
| Freshness signals (dates) | 39017% | 1657% | 1,74076% |
| Offerings marked up | 613% | 00% | 2,23497% |
| Answerability | |||
| Concise, extractable answers | 1,68974% | 28512% | 32114% |
| Question-led content | 22910% | 33315% | 1,73376% |
| Heading structure | 1,18352% | 52923% | 58325% |
| Machine Readability | |||
| Substantive content | 1,78378% | 28713% | 22510% |
| Title quality | 1,85481% | 40017% | 412% |
| Meta description quality | 1,59469% | 44019% | 26111% |
| Image alt coverage | 1,96185% | 1456% | 1898% |
| Semantic structure | 2,20896% | 874% | 00% |
| AI Access & Discovery | |||
| llms.txt (forward-looking) | 58425% | 1,71175% | 00% |
| AI crawler access | 2,25498% | 161% | 251% |
| Indexable & snippet-eligible | 2,26199% | 00% | 341% |
| Canonical URL | 1,96285% | 33315% | 00% |
| Sitemap | 1,24954% | 91240% | 1346% |
| Social / Open Graph | 88939% | 82036% | 58626% |
| Mobile, language & HTTPS | 2,17595% | 1135% | 7under 1% |
Agent Readiness
Can an automated agent get in at the edge and find anything machine-operable when it arrives.
| Check | Pass | Partial | Fail |
|---|---|---|---|
| Agent Access | |||
| AI user-agent treatment at the edge | 2,01688% | 26211% | 171% |
| Robots policy for retrieval agents | 2,25498% | 161% | 251% |
| Actionability | |||
| Machine-readable action | 56625% | 412% | 1,68874% |
| Offerings described | 613% | 00% | 2,23497% |
| Operability | |||
| Form / task operability | 90339% | 1,39261% | 00% |
| Static content (render dependency) | 2,02288% | 1356% | 1386% |
| Machine-readable contact | 78834% | 1,50766% | 00% |
| No CAPTCHA wall | 2,11592% | 944% | 864% |
| Discoverability | |||
| Declared machine interface | 45020% | 1,84580% | 00% |
| Discoverable sitemap | 2,13993% | 00% | 1567% |
Brand Representation
Can a machine confirm from the site itself which real-world organisation it is.
| Check | Pass | Partial | Fail |
|---|---|---|---|
| Identity | |||
| Entity definition | 1707% | 89539% | 1,23054% |
| Identity consistency | 291% | 98643% | 1,28056% |
| Disambiguation | |||
| Authoritative profiles linked | 29813% | 35315% | 1,64472% |
| Spans registry and profile | 1286% | 00% | 2,16794% |
| Authority | |||
| Credentialed authorship | 492% | 1185% | 2,12893% |
| Citation-worthy claims | 84737% | 61327% | 83536% |
| Sourced claims (outbound citations) | 8under 1% | 402% | 2,24798% |
| Representation Rights | |||
| Representation & usage rights | 2159% | 2,08091% | 00% |
What the gaps mean for a business
An AI system that answers a buyer's question has to decide which business it is talking about, what that business sells and whether to trust it. The commonest gaps above sit exactly there: no machine-readable statement of who the business is, nothing marked up about what it sells, and no named, credentialed person behind the content. All three are closed in the site's code rather than with new content. That closing them changes how often an AI system names a business is an inference, and it is not something this scan measures.